Because you motivated us with Smolin's three roads to quantum gravity,
let me tell you that, modulo vocabulary, I tend to be 100% ok with
what Smolin summarizes in his chapter 3 "many observers not many worlds",
including his "psychological" move with the reference of toposes in
And that move is certainly coherent with respect to the comp hyp
(from wich physics is made into a branch of (machine) psychology.
That psychology can itself be embodied in
number theory, computer science, information theory, deformation
theory, etc. (with some "forgetful functor" let us say).
But when Smolin says "many observers not many worlds" he is
somehow playing with words. His own explanation makes clear that
collapse does not occur and that the many observers must come
from realtive differentiation of macro-observers.
An expression like "many observers not many worlds" is ambiguous.
If you take it as "many subjective worlds and one objective world"
then Smolin just contradicts himself, especially through his topos
move. If you take it as "many subjective worlds no objective world"
you get what the comp picture predicts.
If you take it as "many subjective worlds and one objective
sort of superunuiverse or multiverse, well, you get quite standard
form of MWI ...
In MWI, observation is just interaction (tensorial product).